On Fri, 11 Sep 2009, [email protected] wrote:

> Ha! What happens if something else gets killed instead (sshd? iptables?
> syslogd?)...then things get really ugly...not only is apache running badly,
> but you can't login, or the machine is vulnerable, or you loose any logging of
> what processes got killed by the OOM killer. I've had each of those things
> happen, and it's not pretty.

Exactly.  Once the OOM killer activates the system is in an indeterminant 
state, which is highly undesirable.

If swap is enabled on a web server in a farm and it starts to thrash the 
box can be taken out of the pool by the load balancer on the basis of poor 
response times.   Once the box recovers it can return dynamically.

> I like the airbag analogy posted earlier...you never, ever want to use swap

Yes I liked this analogy too.

The nature of swap has changed.  In decades past it was expected that most 
systems would swap a little during normal operation.  These days that 
doesn't need to be the case but swap still has role to play.

> space...but it's better than the alternative. Any use of swap is a 
> signal to buy more RAM or find out the underlying problem. Not having 
> the swap in the first place almost certainly means that a medium-sized 
> problem becomes a really big problem.

Well said.

Rob

-- 
I tried to change the world but they had a no-return policy
http://www.practicalsysadmin.com
_______________________________________________
Discuss mailing list
[email protected]
http://lopsa.org/cgi-bin/mailman/listinfo/discuss
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to